Sounds like weak IV to me; x4 and x5 may not strongly affect y2.
Check with an equivalent linear IV model using -ivreg2- (on SSC).
On Mon, Aug 3, 2009 at 4:31 PM, Martin Weiss<[email protected]> wrote:
>
> <>
>
> Sorry to say, your suspicion was well founded as -ivprobit- is not designed
> for binary endogenous variables, as I just learned from the manual. You may
> want to look at -ssc d cmp- which allows you to emulate the results of
> -biprobit-, but possibly with more flexibility...
>
>
> HTH
> Martin
>
>
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of Sun, Yan (IFPRI)
> Sent: Montag, 3. August 2009 19:54
> To: [email protected]
> Subject: st: Explanation of endogenous binary IV using "biprobit"
>
> Thank you, Martin, for your reply. Could you be more specific about what you
> mean "This sounds more like a problem for -ivprobit"? I think "ivprobit" is
> for binary dependent variable and continuous endogenous variable and
> "biprobit" is for binary dependent variable and binary endogenous variable.
> Am I using the correct command? Thanks. Yan
>
> *********************************************
> This sounds more like a problem for -ivprobit-, though. Be sure to check it
> out before you carry on with -biprobit-...
>
>
>
>
> HTH
> Martin
>
>
> -----Ursprüngliche Nachricht-----
> Von: [email protected]
> [mailto:[email protected]] Im Auftrag von Sun, Yan
> (IFPRI)
> Gesendet: Montag, 3. August 2009 17:45
> An: [email protected]
> Betreff: st: Explanation of endogenous binary IV using "biprobit"
>
> Hi,
>
> I have the following equations
> Y1=Y2 X1 X2 X3 (1)
> Y2=X1 X4 X5 (2)
>
> Where Y1 is binary dependent variable and Y2 is binary endogenous
> variable, Xs are exogenous variables.
>
> When I run Probit model of equation (1), Y2 is positive significant; but
> when I run equations (1) and (2) together using biprobit, Y2 is not
> significant at all (in the 2nd stage), under the wald test of rho: chi2
> is very small (with prob very big), which implying rho is not
> significant and equation (1) and equation (2) is not correlated,
> although X4 and X5 are significant. What does it mean? Does it mean that
> Y2 is not endogenous variable and I could use probit model for equation
> (1)? Does it also mean that Y2 has a significant impact for Y1?
>
> Thank you very much.
>
> Yan
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
